[github-actions] Retry transient workflow failures (#175)

[coisa]

Related Issue

Closes #175

Motivation / Context

Repository workflows occasionally fail because GitHub-side infrastructure flakes during checkout or fetch operations even though rerunning the same run immediately succeeds. We want a bounded, visible rerun policy for those transient failures without teaching CI to paper over real workflow, test, or validation regressions.

Changes

• add a repository-local retry-transient-failures.yml workflow that listens for failed workflow runs
• inspect failed job logs for transient GitHub-side checkout and transport signatures such as HTTP 500 fetch failures and packfile or ref-listing transport errors
• rerun failed jobs once when every failed job matches the transient signature set
• emit a deterministic step summary showing whether the rerun was requested or skipped
• document the retry behavior in the README and GitHub Actions usage guide
• add a changelog entry for the new workflow resilience behavior

Verification

• ruby -e 'require "yaml"; YAML.load_file(".github/workflows/retry-transient-failures.yml"); puts "ok"'
• composer dev-tools changelog:check
• git diff --check

Documentation / Generated Output

• updated README.md
• updated docs/usage/github-actions.rst
• updated CHANGELOG.md

Changelog

• Added an Unreleased changelog entry for transient GitHub Actions retry behavior.

Reviewer Notes

• .github/wiki remains locally modified in the workspace but is intentionally not part of this PR.
• This implementation is repository-local on purpose. It improves this repository's workflow resilience without changing the packaged consumer workflow set yet.

[github-actions]

🚀 Preview is available for this pull request.

• Docs: https://php-fast-forward.github.io/dev-tools/previews/pr-176/
• Coverage: https://php-fast-forward.github.io/dev-tools/previews/pr-176/coverage/
• Metrics: https://php-fast-forward.github.io/dev-tools/previews/pr-176/metrics/